Did the Bank of Japan's negative interest rate policy increase bank lending?

After Denmarks Nationalbank temporarily imposed a negative interest rate on central bank deposits in 2012, similar policies were adopted by the European Central Bank (ECB) and the Swiss National Bank in 2014, the National Bank of Sweden in 2015, the Swedish National Bank in 2015, and the Bank of Japan in 2016. Gunji and Miyazaki (2016) show that the negative interest rate policy on central bank deposits has a negative impact on bank lending and deposits, using a theoretical model that applied the Cournot model to the banking industry. This is contrary to the expectation of the central banks of countries that have introduced the negative interest rate policy so far.

This paper examines whether the negative interest rate policy adopted by the Bank of Japan in 2016 affected bank lending by using regression discontinuity design (RDD) using bank level data. The Bank of Japan announced the Amendment to “Condition of Complementary Deposit Facility as a Temporary Measure to Facilitate Supplying of Fund” at the Monetary Policy Meeting on 29th January 2016, and implemented a negative interest rate policy on February 16, 2016. The introduction of the negative interest rate policy of the Bank of Japan was not expected at all until the Monetary Policy Meeting in January 2016. In fact, Haruhiko Kuroda, the President of the Bank of Japan denied the adoption of the negative interest rate policy until just before that, and many market participants believed it. Therefore, the introduction of the negative interest rate policy of the Bank of Japan can be treated as a natural experiment in verifying its effect. Moreover, since negative interest rates are not imposed on all banks but only on banks that satisfy certain conditions, it is possible to verify policy effects by comparing them.

We find from the empirical results that the negative interest rate policy reduced bank lending by 1.5% to 3.5%. This result was not changed by parametric estimation or nonparametric estimation. In addition, even if the estimation model was made linear or quadratic function, the result was not affected. On the other hand, similar estimates were made on the rates of change in deposits, securities and call loans, but no clear conclusion was obtained.